Figure out a problem or pain-point that you experience in everyday life. If something is annoying or frustrating, chances are that there is a better way to solv

Figure out a problem or pain-point that you experience in everyday life. If something is annoying or frustrating, chances are that there is a better way to solve it. Look for assumptions that might no longer hold. For example: traditional datastructures like B-trees assume hard-disk-based storage where random access times are slow. But what if we convert everything to flash SSDs or run things in RAM, does that open up a space for innovation (eg: storage systems based on log-structured merge trees and other more exotic datastructures)? Try juxtaposing 2 things that are not normally considered to be related, and then combine them into a mashup. Start with a small, simple prototype testing the core of your thesis, then iteratively improve it and scale it up. Don't try to boil the ocean, you'll just end up frustrated.